Neutron activation analysis and data mining techniques to discriminate between beef cattle diets

Abstract

Neutron activation analysis and data mining techniques were combined for assessing the mineral composition of diets commonly used to feed beef cattle in Brazil. Among twenty chemical elements determined, Br, Ca, Cs, La, Sc, Se, Sr, Th and Zn showed statistically significant differences between the two cattle diets studied. Chi square indicated that Cs, Se and Sc provided better diets discrimination. The highest classification performances using these elements were achieved for multilayer perceptron and sequential minimal optimization with prediction accuracy of 100%.